Analysis of the Relationship between Managers Compensation and Earnings in Companies Listed in the Tehran Stock Exchange

Transcription

1 ORIGINAL ARTICLE Receved 05 Jul. 03 Accepted 05 Aug , Scence-Lne Publcaton ISSN: Journal of Educatonal and Management Studes J. Educ. Manage. Stud., 3( , 03 JEMS Analyss of the Relatonshp between Managers Compensaton and Earnngs n Companes Lsted n the Tehran Stock Exchange Bjan Abedn, Mohammad Hossen Ranjbar and Ahmad Sanjar Majan * Department of Accountng, Islamc Azad Unversty, Scence and Research Branch, Hormozgan, Iran * Correspondng author s Emal: salarx7000@yahoo.com ABSTRACT: The objectve of the present study s to study the relatonshp between managers compensaton and earnngs n companes. It s a causal-analytc lbrary survey whch s based on the analyss of panel data. In ths research the fnancal nformaton on companes lsted n the Tehran Stock Exchange from 006 to 00 s analyzed. It shall be mentoned that 560 companes were lsted n the Tehran Stock Exchange each year. The SPSS 0, EVIEWS 7 and MINITAB 6 software were used for analyzng the research results. Results of the study of the research hypothess ndcated that there s a sgnfcant drect relatonshp between managers compensaton and earnngs. Key words: Managers Compensaton, Earnngs Fluctuatons, Panel Data. INTRODUCTION From the late 980s to the early 990s, most bg nternatonal companes allowed ther top managers to buy actons at the base prce and have a share of the earnngs f ther performance would contrbute to the growth of the company. For about one decade the method was popular, because the securtes market was bloomng for some subsequent years. Therefore, managers and stockholders benefted a great deal from the market, but after the fall of stock markets n 000 bg companes fgured that the method does not always yeld satsfactory results. Hence, the relatonshp between managers compensaton and the earnngs s studed n ths research. Research Lterature Foregn Researchers: Natarajan (99) carred out a study ttled Applcaton of Optonal Accountng Reports to Managers Compensaton Contracts. In hs study he analyzed tools for measurng the effect of compensaton on the performance of managers. One of the man ways of measurng the performance of the management s by usng accountng tools. Snce the measurng tools gve nformaton on the performance of managers and nvestment decsons, they are used as means of measurng and assessng the performance and motvatons of managers. Arya et al. (008) conducted a study ttled Do stockholders always prefer managed earnngs? At the end of the study they found out that managers remove the unstable components of earnngs through ncome smoothng and establsh a seral connecton between dfferent annual earnngs. As a result, estmatons of captal markets about actons potentally depend on the ncome smoothng patterns used by managers. Therefore, at tmes when smoothng chances are decreased, unusual varatons of market value and corporate return are observed. Shalev et al. (0) n ther study, Accountng Income and Top-Management Compensaton, analyzed the effect of the structure of managers compensaton on the current values of companes resulted from ntegraton. They clamed that managers compensaton may change the results of assessments and nformaton revealed on assets and non-monetary labltes. Results of the research showed that the structure of managers compensaton s an mportant factor n determnng the value of the obtaned goodwll. The reason s that the goodwll n purchased companes s an mportant factor n determnng expected earnngs and can also nfluence the current and future managers compensaton. In addton, results of the research ndcated that snce the performance and effcency of companes are determned based on varous accountng crtera, compensaton contracts affect the accountng methods employed by managers. Namaz and Sran (004) performed a study ttled Emprcal Analyss of Factors Important n Determnng Contracts, Indcators, and Compensaton Parameters for CEOs of Iranan Companes. They wdely studed managers contracts of companes lsted n the Tehran Stock Exchange. Results of the study are as follows: ) On an average bass there s a conflct between the nterests of managers and stockholders; ) A contract whch makes a balance between the earnngs of stockholders, managers and employees contrbutes to the enhancement of To cte ths paper: Abedn B., Ranjbar MH. And Sanjar Majan A. 03. Analyss of the Relatonshp between Managers Compensaton and Earnngs n Companes Lsted n the Tehran Stock Exchange. J. Educ. Manage. Stud., 3(

2 Abedn et al., 03 effcency; 3) A contract whch s only based on accountng crtera and does not take factors such as prce and corporate return nto account rarely leads to an ncrease n value. Thaqaf and Kurdestan (004) conducted a study ttled Analyss and Clarfcaton of the Relatonshp between Earnngs Qualty and Market s Response to Varatons of Dvdend. The sample under study ncluded 50 companes actve from 99 to 00. Results of the research suggest that accordng to the qualty of earnngs measured based on earnngs persstence, the market shows a postve reacton to the expected corporate return and corporate return s not affected by the accrual components of earnngs. Nourvash and Hussen (009) carred out a study ttled Analyss of the Relatonshp between Qualty of Dsclosure (Relablty and Tmelness) and Earnngs Management. In ths study, qualty of dsclosure was measured usng the followng two crtera: tmelness and relablty. Earnngs management n companes was also assessed usng the modfed Jones model. Fndngs of the research ndcate that there s a negatve sgnfcant relatonshp between the qualty of corporate dsclosure and earnngs management. Moreover, t was also found out that there s a negatve sgnfcant relatonshp between tmelness of corporate dsclosure and earnngs management. Problem Statement Wth the development of commerce and ndustry as well as the emergence of gaps between owners and managers of enterprses, control and compensaton mechansms have found to be necessary. Accordng to the agency theory, an organzaton s a combnaton of contracts. The entty of every busness unt depends on ts contracts. The contracts can ether be wrtten or non-wrtten. One of the most mportant types of contracts s compensaton contract for major stockholders (owners) and managers. Agency s a form of contract by whch the employer/owner assgns an agent the task of decson-makng (Namaz, 006). The rato of compensaton to earnngs s determned by nvestment opportuntes, earnngs persstence, and earnngs response coeffcent (Bushman et al., 006). Accordngly, the man queston rased n ths research s whether there s a relatonshp between managers compensaton and earnngs. The objectve of the present study s to analyze the relatonshp between the managers compensaton and earnngs n companes lsted n the Tehran Stock Exchange. Hence, the scentfc and practcal objectves of the research are determned as follows: ) Determnng the relatonshp between managers compensaton and earnngs n companes lsted n the Tehran Stock Exchange. ) Clarfyng the relatonshp between performance crtera and managers compensaton. Moreover, the research hypothess also states that there s a relatonshp between managers compensaton and earnngs. MATERIALS AND METHODS The statstcal populaton for the present research ncludes all of the companes lsted n the Tehran Stock Exchange. Accordng to the offcal webste of the Tehran Stock Exchange, a total of 458 companes from 37 ndustres had been lsted n the Tehran Stock Exchange by 0 Therefore, all of the companes lsted n the Tehran Stock Exchange from 006 to 00 formed the statstcal populaton for the present study. In order for the sample to be a true representatve of the statstcal populaton, samples were selected by the Crtera-Flterng Technque. The followng crtera were met for choosng the samples. That s to say, each of the sample companes meets all of the samplng crtera, whch are as follows: - Accordng to the nformaton requred from year 005, all of the companes that were lsted n the Tehran Stock Exchange as of March, 006 and those companes whose names were stll lsted n the Tehran Stock Exchange by 0 - Companes whose stock was exchanged actvely n the Tehran Stock Exchange durng the aforementoned perod. 3- Those companes whose fscal perod ends to 0th March of each year and dd not experence changes of fscal perod durng the aforementoned perod. 4- Companes that do not belong to brokers (nvestment companes, holdng, leasng, banks and nsurance companes). 5- In order to calculate market value at the end of each fscal year by Tobn s Q rato, only companes are selected whose stock s exchanged at least once from January to February 0 or from February 0 to March each year. 6- Companes wth requred nformaton avalable. The crtera were checked by the order of appearance. That s to say, f a company met the frst crtera, t was not checked to see whether t met the next crtera. Research Samples A total of companes from 9 ndustres were found to meet the crtera. The samples were used to analyze research hypothess. Therefore, the sample sze s not equal to the sze of the statstcal populaton. The present study s a descrptve study for the method t employs for collectng nformaton. Based on ts objectve t s also a developmental research and 456

3 J. Educ. Manage. Stud., 3( , 03 based on ts methodology t s a correlaton research. It apples the aforementoned theores to the newly emerged Iranan market. Frst, data and nformaton assocated wth research varables were extracted manually from accountng tems contaned n the audted fnancal statements of companes lsted n: the Research, Development and Islamc Studes webste afflated wth the Stock Exchange at the CODAL network or the comprehensve system of publshers nformaton at Iranan Center for Fnancal Informaton Processng at and compact dscs publshed by the Stock Exchange. Addtonal nformaton on the fnancal statements of companes, whch s stored n the Stock Exchange database, was also extracted usng Novn Rah Avard software n pdf and.xls formats. The data was also analyzed usng Excel, SPSS, and EVIEWS software. Defnton and Calculaton of Varables Dependent Varables: Managers Compensaton Contracts ( ΔLnComp, ) Managers Compensaton: Organzaton usually rewards managers n order to compensate for the creatvty and nnovatons of managers n fndng and employng new and better procedures and methods. Such rewards are normally gven to managers that accomplsh tasks wth a qualty hgher than the expected qualty (Namaz, 006). The reward or compensaton value s determned based on audted balance sheets. Independent Varables Earnngs ( EARN, t Accountng Income: Accountng ncome or net proft refers to the dfference between the ncome resultng from sales of goods and manufacturng servces of a regstered enterprse and the total regstered expenses of the enterprse. In ths research, the earnngs value obtaned wthout extraordnary tems s used as a crteron for assessng executve performance. Control Varables Corporate Total Return ( RET, Total Return: Total return refers to the combnaton of nterests assocated wth one share durng a perod of tme. The perod can be a day, month, or year. The benefts nclude cash payments such as dvdend and the dfferent between stock s prce at the begnnng and at the end of the perod of concern. In the present study, return s used as a crteron for assessng the performance of the management. The rato varatons of earnngs to varatons of returns ( ( earn) / V( ret) V, t t t The rato of the varatons of earnng to varatons of return s calculated by dvdng varaton of earnngs by varaton of return. Investment opportuntes (rato of market value to book value) ( MB, t Investment opportuntes are equal to the rato of market value to book value (MB) of captals and are calculated usng the followng relaton: MTB, t = Corporate market value s obtaned by multplyng the mean stock market prce durng the fscal perod by the number of ssued shares. Book value s also the book value of equtes at the end of the fscal year. Earnngs Persstence ( PERSIST, Earnngs Persstence: It refers to the contnuty (repeatablty) of current earnngs. Increase n the earnngs persstence shows that the company s more capable of keepng current earnngs and has earnngs wth more qualty. In ths research, earnngs persstence s measured usng the Lpe-Kormend regresson model (987). Accordng to the Lpe-Kormend regresson model (987) earnngs persstence s measured by regresson of current earnngs on prevous perods earnngs. EARN, t =β 0 + β EARN, t- +, t In the above regresson model, coeffcent reflects earnngs persstence. In the test model developed for research hypotheses the varable s shown by PERSIST. Earnngs response coeffcent ( ERC, Earnng Response Coeffcent: Theoretcally, t refers to a state where the nformaton presented on the captal market can nfluence the decsons of useful nvestors. In order to measure the varable, the earnng response coeffcent (ERC) s used. The value of the coeffcent s obtaned by regresson of net proft on return. In addton, n order to measure the earnngs value relevance based on the method by Bushman et al. (006) the earnng response coeffcent s also used. The coeffcent s obtaned from the followng regresson model: RET, t =β 0 + β EARN, t +, t In the above regresson model, s the earnng response coeffcent, whch reflects the earnngs response coeffcent? In the test model for the hypotheses the varable s shown by ERC (Peng, 0). The Method for Analyss of Models and Examnaton of Research Hypothess: In ths research the hypothess under examnaton s studed usng the multvarate lnear regresson t t 457

4 Abedn et al., 03 model. Frst, the statstcal method used n the research, whch s the panel data method, s descrbed. Second, tests performed for determnng the sgnfcance of the model and ts ndependent varables are explaned. Each statstc explans how to reject or approve the results of the tests. Research hypotheses were examned usng SPSS (ver. ), EVIEWS (ver. 7) and MINITAB (ver. 6) software were employed. Panel Data Method In general, the followng model s a model that works wth panel data: k Y t X t k kt Fxed Effects Method kt The fxed effects method assumes that varable coeffcents (slopes) are fxed and the dfferences between unts can be expressed by y-ntercept dfferences. If dfferent y-ntercept values are determned for dfferent sectonal unts, the so called One-Way Fxed Effects method s used. The model for the method s expressed as follows: Y t X k k k kt However, f the y-ntercept values of sectons and perods vary from each other, the Two-Way Fxed Effect method s used. The model for ths method s as follows: Y t X t k k k t t kt Chow Test or Bound F Test In order to determne whether panel data or combned data are useful for estmatng the desred functon, a hypothess s examned whch states that all of the fxed terms of the estmaton are equal. The null hypothess for ths test, whch s known as the Chow test or the bound F test, s as follows: H H 0 : : In order to examne the aforementoned hypothess, the F statstc s used as follows: ( RRSS URSS) /( N ) F( N, NT N K) URSS /( NT N K) Hausman test When results of the Chow tests performed on each hypothess confrm the use of panel data, n order to determne the best estmaton method (fxed or random effects of the dfferences between sectonal t unts) the Hausman test method s employed. The statstc for the test s obtaned as follows: The null hypothess for the Hausman test s as follows: H0 : E( ut X t ) 0 H : E( ut X t ) 0 Model Sgnfcance Test In order to determne the sgnfcance of the regresson model the F statstc was used. The null hypothess for the F test s as follows: H 0 :... k 0 H :... k 0 The accuracy of the hypothess s examned usng the followng statstc: ESS /( K ) F RSS /( N k) Gude: When the calculated F s larger than the desgn F ( F F ( K, NK ) ), The numercal value of the test functon fall nto the crtcal area and the null hypothess ( H 0 ) s rejected. In ths case, the entre model s sgnfcant at a confdence level of 95%. Coeffcents Sgnfcance Test In order to study the sgnfcance of coeffcents of ndependent varables of each model the t statstc s used. The null hypothess for the t-test s as follows: H H 0 : 0 : 0 The accuracy of the hypothess s examned usng the followng statstc: ˆ T SE( ˆ ) ~ t, N k When the null hypothess ( H 0 ) s rejected, the coeffcent of concern ( ) wll be sgnfcant at a confdence level of 95%. Postvty of the coeffcent also mples that there s a relatonshp between the ndependent and dependent varables. Examnaton of Normalty of Varables Dstrbutons In order to examne the normalty of the varables the Kolmogorov-Smrnov test was put nto use. The null hypothess and statstc for the test are as follows: H H 0 : Normal Dstrbuton : Not Normal Dstrbuton 458

5 J. Educ. Manage. Stud., 3( , 03 D Max F, N N N Y FY Gude: When the probablty of the test s larger than 05, t s concluded that the varables and remans are dstrbuted normally (confdence level=95%). Examnaton of Hypotheses of the Lnear Regresson Model In order to choose the Best Lnear Unbased Estmators (BLUEs) as the ordnary least squares of regresson coeffcents, the followng hypotheses are defned for the mode: - The remans are normal: One of the hypotheses of the regresson model s that zero-mean errors are normally dstrbuted. To prove the hypothess, the mean value of data should be small and close to zero. Its standard devaton should also be close to one. SPSS software s used for performng the test and creatng the assocated dagrams. In addton, the Kolmogorov-Smrnov test, as a nonparametrc test, s performed to examne the normalty of the remans. The statstc for the test can also be calculated usng SPSS software. Gude: When the value of the statstc obtaned from the test s larger than 5%, the null hypothess s approved. The hypothess states that the varable under study s normally dstrbuted at a confdence level of 95%. - Independent varables lack collnearty: Collnearty refers to the lnear relatonshp between explanatory and ndependent varables. One of the ways of ndentfyng collnearty or lack of collnearty s to analyze the correlatons among ndependent varables. If there s not a severe correlaton among ndependent varables, collnearty wll not occur. In ths study, the collnearty of ndependent varables s studed usng Pearson s correlaton coeffcent. - Remans are ndependent: One of the hypotheses of regresson analyss s ndependence of errors (or the dfference between true values and values predcted by the regresson model). If the ndependence of errors s not proved and errors have correlatons, regresson models cannot be used. In order to study the ndependence of errors the Durbn- Watson test s employed. The statstc for ths test falls nto the range of If the statstc falls nto the range of.5-.5, the null hypothess, H 0, (whch stands for lack of correlaton among errors) s approved and the regresson method can be used. The test can be performed usng SPSS software (Mo men and Fa al Qayum, 007). - There s no heteroscedastcty among the remans: The Lagrange Multpler (LM) s used to test ntergroup heteroscedastcty usng panel data. The statstc can be calculated by performng an overall OLS on the model usng combned data. LM T S ( ) ~ X ( n) S Gude: When the value of the calculated statstc s smaller than the crtcal value at a confdence level of 95%, the H 0 hypothess s approved and the heteroscedastcty among sectonal unts s rejected. - There s no msspecfcaton and lnearty n the model Neglectng varables that should be contaned n the equaton (due to lack of awareness about ther exstence, unavalablty of relevant nformaton, addng undue varables to the equaton, an ncorrect functon form (e.g. choosng a lnear functon nstead of a logarthmc functon), etc. can lead to model msspecfcaton. In ths research, n order to study model msspecfcaton the Ramsey model s used. The statstcal hypothess for ths test s expressed as follows: The stages of the test are as follows: ) Obtanng estmated Y values (Yˆ s) ) Estmaton of the model as follows: Y X Yˆ Yˆ 3 e 3) Calculaton of the test statstc as follows: RNew ROld P F ( RNew) ( n P ) 4) If the calculated F s larger than the desgn F, the model has msspecfcaton. RESULTS In ths research, data s analyzed usng central ndcators such as average and mean values as well as dsperson ndcators of standard devaton. However, average s the man central ndcator, whch shows the average value of the data. In addton, standard devaton s among the dsperson parameters that shows the level of dsperson of the data. Skewness s also one of the parameters showng devatons from symmetry and s known as the symmetry ndcator (Mo men and Qayum, 0). Partculars of the descrptve statstcs on the model varables are shown n Table. The statstcs are obtaned usng the crtera-flterng technque and by removng the outlers n SPSS software. 459

6 Abedn et al., 03 Varable Table. Descrptve statstcs on the research varables Number of observatons Average Standard Devaton Mnmum Maxmum Skewness Kurtoss Managers compensaton Earnngs Total return Varance of the rato of earnngs to return Rato of market value to book value Earnng persstence Earnngs response coeffcent Accordng to Table, the average value of the managers compensaton, as an ndependent varable, s The varable should have values of 0 and 3 n order to have a normal dstrbuton. Analyss of the aforementoned values ndcates that managers compensaton s a varable wthout normal dstrbuton. Accordng to the descrptve statstcs presented n Table, the average value of the ndependent varable, earnngs of sample companes, s equal to for the perod under study. The varable should have values of 0 and 3 n order to have a normal dstrbuton. Analyss of the aforementoned values ndcates that earnngs level s a varable wth normal dstrbuton. Moreover, the postve average values of control varables ncludng corporate total return, the rato of the varatons of earnng to varatons of return, rato of market value to book value, earnngs persstence and earnngs response coeffcent are 0348, 345, 809, 973, and 9464, respectvely. Examnaton of Normalty of Dstrbuton of Research Dependent Varables Snce the sgnfcance of the K-S statstc for managers compensaton s less than 05, the H0 hypothess, whch states that the dstrbuton of the varables at a confdence level of 95% s normal, s rejected. Therefore, managers compensaton s a varable wthout normal dstrbuton. Table. Results of examnaton of normalty of the research dependent varable Varable Number (N) (K-S) statstc Sgnfcance level (sg) Managers compensaton Dagram. The non-normal dstrbuton of managers compensaton 460

7 Abedn et al., 03 Dagram (Q-Q plot) shows the non-normal dstrbuton of the dependent varable. Normalty of the dependent varable s the prerequste for regresson models. Therefore, before testng the hypotheses the varable (s) should be normalzed. In ths research the data s normalzed usng Johnson s Transformaton Functon. The analyses are also performed usng MINITAB (ver. 6) software. The results obtaned from the K-S test performed followng to normalzaton of the data are llustrated n Table 3. Table 3. Results of examnaton of the normalty of the research dependent varable after the normalzaton process Varable Number (N) (K-S) statstc Sgnfcance level (sg) Managers compensaton Accordng to Table 3, snce after data normalzaton the sgnfcance level (sg.) of the Kolmogorov-Smrnov statstc for managers compensaton s more than 05 (497), the H0 hypothess s approved at a confdence level of 95%. Ths reflects the fact that managers compensaton s a varable wth normal dstrbuton followng to the normalzaton process. Dagram (Q-Q plot) shows the normal dstrbuton of dependent varables after normalzng the data usng Johnson s Transformaton Functon. Dagram. The normal dstrbuton of managers compensaton followng to the normalzaton process Analyss of Correlatons among Research Varables In ths secton, usng Pearson s correlaton coeffcent the relatonshps or correlatons between research varables s studed. The correlaton coeffcents matrx for the research varable s shown n Table 4. Accordng to the results obtaned usng Pearson s correlaton coeffcent, there s a postve sgnfcant correlaton between corporate earnngs and total returns. In addton, there s also a postve sgnfcant relatonshp between earnng persstence and rato of market value to book value. Analyss of Collnearty of Research Varables In ths secton n order to examne the research hypothess an estmaton model s determned at the begnnng. Next, the research model s estmated and results are nterpreted. Moreover, statstcal hypotheses for testng the research hypothess are also analyzed and presented. The hypotheses nclude normalty of remans, homoscedastcty n remans, ndependence of remans, and lnearty of the model. Results of Testng the Research Hypothess The objectve of the research hypothess s to study the relatonshp between managers compensaton and earnngs. The statstcal hypothess for the research hypothess s expressed as follows: H0: There s no sgnfcant relatonshp between managers compensaton and earnngs. H: There s a sgnfcant relatonshp between managers compensaton and earnngs. 38

8 Abedn et al., 03 Managers compensaton ( P Value) Earnng ( P Value) Corporate total return ( P Value) Rato of the varatons of earnng to varatons of return ( P Value) Rato of market value to book value ( P Value) Earnng persstence ( P Value) Earnngs response coeffcent ( P Value) Table 4. Pearson s correlaton coeffcents matrx for research varables Managers compensaton 000 (9) 000 (9) 00 (008) 007 (60) 060 (00) 009 (80) Earnngs.000 (000) 000 (998) 00 (8) -009 (353) -0 (770) Corporate total return 000 (998) 00 (8) -009 (000) -0 (770) Rato of the varatons of earnng to varatons of return 006 (09) -00 (06) 000 (000) Rato of market value to book value 09 (09) 0 (60) Earnng persstence 00 (80) Earnngs response coeffcent The hypothess s examned usng the followng model and panel data. If the value of the coeffcent at a confdence level of 95% s sgnfcant the hypothess s approved. In ths model: I: the company (sectonal unts); T: year; ΔLnComp: varatons of compensatons for the board of drectors durng perod t compared to perod t-; ΔEARN: varatons of accountng ncome durng perod t compared to perod t-; RET: corporate total returns V (earn)/v (ret rato of the varaton of earnngs to varaton of returns durng the course of the research; MB: rato of market value to book value; PERSIST: earnng persstence; ERC: earnngs response coeffcent; Corporate random error n year t; DISCUSSION Results of the Chow and Hausman tests on the research model are presented n Table 5. It shall be mentoned that the former test s performed to choose between panel data or combned data methods whle the latter one s performed for dentfyng the effects of the fxed effects method or the random effects method on the panel data. Accordng to the results of the Chow test and ts p- value (063), the H0 hypothess s rejected at a confdence level of 95%. Hence, the panel data method s a sutable method for ths model. Moreover, accordng to the results of the Hausman test and ts p- value (009), whch s less than 05, t can be sad that the H0 hypothess s rejected at a confdence level of 95% whle the H hypothess s approved. Therefore, the model shall be estmated usng the fxed effects method. Results of the Jarque-Bera test performed on the classc regresson hypotheses suggest that remans of estmaton of the research model at the confdence level of 95% are normally dstrbuted. Therefore, the probablty of the test (547) s bgger than 05. In addton, accordng to the sgnfcance level of the Breusch-Pagan test, whch s bgger than 05 (065) t can be sad that the null hypothess, whch supports homoscedastcty, s approved. Hence, the problem s not nvolved wth heteroscedastcty. In the autocorrelaton test, reman of the model, whch s tested usng the Durbn-Watson statstc, s equal to.47. Snce the value falls between.5 and.5, t can be concluded that the remans are ndependent. Moreover, snce the sgnfcance level of the Ramsey test s bgger than 05 (734), the null hypothess (lnearty of the model) s approved and the model s not nvolved wth msspecfcaton. Partculars of the results of the aforementoned tests are presented n Table 6. Table 5. Results of the Chow and Hausman tests performed on the research model Test Statstc Statstc value Degree of freedom P-Value Chow F 009. ( 00) 060 Hausman

9 J. Educ. Manage. Stud., 3( , 03 ( P Value) Ramsey statstc Table 6. Results of examnatons of the statstcal hypotheses of the research model F 0090 Durbn-Watson statstc D 07. Breusch-Pagan statstc ( P Value) F 608 Jarque-Bera statstc ( P Value) Accordng to the results of the Chow and Hausman tests as well as results of examnaton of the statstcal classc regresson hypotheses the research model s estmated usng panel data and the fxed effects method. Results of the estmaton are presented n Table 7. Table 7. Results of examnaton of research hypothess usng the fxed effects method Dependent varable: managers compensaton Number of observatons: 560 year-company Varable Coeffcent t statstc P-Value Relatonshp Fxed component Negatve Earnngs Postve Earnngs vs. varaton of earnngs to Negatve returns rato Earnngs vs. rato of market value to Postve book value Earnngs vs. earnng persstence Insgnfcant Earnngs vs. earnng response Postve coeffcent Corporate total return Insgnfcant Total return vs. varaton of rato of Insgnfcant earnngs to returns Total return vs. rato of market value to Postve book value (MB) Total return vs. earnng persstence Insgnfcant Total return vs. earnng response Insgnfcant coeffcent Varaton of the rato of earnngs to Postve returns Rato of market value to book value Insgnfcant Earnng persstence Insgnfcant Earnng response coeffcent (ERC) Postve Model coeffcent of determnaton 798 F statstc ( P Value).60 (0000) The estmated verson of the model processed n EVIEWS (ver. 7) software s as follows: Snce the probablty value of the F statstc s smaller than 05 (0454), the sgnfcance of the entre model s approved at a confdence level of 95%. The model coeffcent of determnaton also shows that 7.98% of managers compensaton s explaned by varables ntroduced nto the model. Accordng to the results presented n Table 7, snce the probablty value of the t statstc for the varant coeffcent of earnng s smaller than 05 (0453), there s a sgnfcant relatonshp between managers compensaton and earnngs at a confdence level of 95%. Hence, the research hypothess s approved and t can be sad that there s a sgnfcant relatonshp between managers compensaton and earnngs (confdence level=95%). The postvty of the coeffcent (5488) reveals that there s a drect relatonshp between managers compensaton and earnngs. Hence, a unt ncrease n the earnngs leads to a 5488 unt ncrease n managers compensaton. General Concluson Snce the probablty value of the F statstc s smaller than 05 (0454), the sgnfcance of the entre model s approved at a confdence level of 95%. The model determnaton coeffcent also shows that 7.98% of managers compensaton s explaned by varables ntroduced nto the model. In addton, accordng to the contents of Table 7, snce the probablty value of the t statstc for earnngs varable s less than 05 (0453), there s a sgnfcant relatonshp between managers compensaton and earnngs at a confdence level of 95%. Hence, the research hypothess s approved and t can be clamed that there s a sgnfcant relatonshp between 7

10 Abedn et al., 03 managers compensaton and earnngs (confdence level=95%). Postvty of the assocated coeffcent (5488) suggests that there s a drect relatonshp between managers compensaton and earnngs. Therefore, a unt ncrease n the earnngs leads to a 0088 unt ncrease n the managers compensaton. REFERENCES Arya, A., J.C., Glover. & Sunder, S. (008). Are unmanaged earnngs always better for shareholders? Accountng Horzons 7: -6. Bushman, R.M., Engel, E. & Smth, A. (006). An analyss of the relaton between the stewardshp and valuaton roles of earnngs. Journal of Accountng Research, 44 ( Kormend, R. & Lpe, R. (987). Earnngs nnovatons, earnngs persstence, and stock returns. Journal of Busness, 60 ( Moment, M. & Fa al Qayum, Al. (007). Analyss of statstcal data n SPSS. Ketab-e-No Publcatons. Mo men, M.r & Fa al Qayum, A. (0). Analyss of statstcal data n SPSS. Terhmeh Publcatons. Namaz, M. (006). Costng n servce frms. Assocaton of Iranan Chartered Accountants, Tehran, May - June : 8. Namaz, M. (006). Evoluton of balanced assessment. Journal of Humantes and Socal Scences. Unversty of Mazandaran, sprng, 6. Namaz, M. & Sran, M. (004). Emprcal analyss of structures mportant n contracts, ndcators, and parameters of compensatons for CEOs of Iranan companes. Journal of Accountng Studes and Audtng, summer, 67. Natarajan, R. (99)."The Use of Dscretonary Accountng Reports n Management Compensaton Contracts", Unversty of Pennsylvana. Nourvash. & Hussen. (009). Analyss of the relatonshp between qualty of dsclosure (relablty and tmelness) and earnngs management. Journal of Accountng Studes and Audtng, 55, 009. Peng, E.Y (0). Accruals qualty and the ncentve contractng role of earnngs.j. Account. Publc Polcy 30 (0) Shalev.R. Zhang, I. & Zhang, Y. (0).Yong CEO compensaton and far value accountng: Evdence from purchase prce allocaton. Oln Busness School Washngton Unversty, St. Lous, MO 6330 Thaqaf, A. & Kurdestan, Q. (004). Analyss and clarfcaton of the relatonshp between earnngs qualty and market response to varatons of dvdends. Journal of Accountng Studes and Audtng, 37, autumn,